Shikib Mehri

Shikib Mehri

Director of Research at Contextual AI, managing a team of researchers working on LLMs and agents. Previously an Applied Scientist at Amazon, where I helped found the Alexa LLM post-training team (which became Amazon AGI). PhD from CMU LTI [thesis] (2022) and BSc from UBC (2018). Always looking to meet exceptional researchers and interns — check out our open roles.

Recent: (1) Grounded Language Model reached #1 on the FACTS leaderboard (VentureBeat) (2) LMUnit open-sourced and achieved #1 on RewardBench2 (3) AgentLens launched as a multi-agent evaluation system (4) Post-training recipe featured as a Google Cloud case study

I most enjoy product-driven research — work that enables next-generation user experiences through strong problem formulation. Currently focused on multi-agent systems, continuous learning, memory, agentic evaluation, and context representations. Key areas include:

BSc from UBC (2018) with internships at Meta (shipped first subword neural MT), Microsoft, and Amazon (patent), and 2 years in bioinformatics at BC Children's Hospital (graph-based genome representation). PhD at CMU LTI (2018–2022) on dialog systems [thesis], advised by Dr. Maxine Eskenazi; was also a TA and research mentor throughout both degrees. During my PhD I also interned at Amazon (dialoglue, example-driven '21). After my PhD I was an Applied Scientist at Amazon, where I helped found the Alexa LLM post-training team (led early SFT, RLHF recipe, data collection, and owned RM/eval) which became Amazon AGI. I then joined Contextual AI as a Member of Technical Staff (Jan 2024), was promoted to Technical Lead Manager (May 2024), and then Director of Research (Aug 2025).